On the polymorphemic genesis of some Proto-Quechuan roots: Establishing and interpreting non-random form/meaning correspondences on the basis of a cross-linguistic polysemy network

In the Proto-Quechuan lexicon, many two-segment phonetic substrings recur in semantically related roots, even though they are not independent morphemes. Such elements may have been morphemes before the Proto-Quechuan stage (i.e., in Pre-Proto-Quechuan). On the other hand, this may simply be due to chance, or to phonesthesia. In this paper, we introduce the Crosslinguistic Colexification Network Clustering (CCNC) algorithm, as well as an accompanying test statistic, which allow us to evaluate our claims against a neutral standard of semantic relatedness (the CLICS2 database; List et al. 2018). We obtain very strong statistical evidence that there are hitherto unexplained recurrent elements within Proto-Quechuan roots, but not within roots reconstructed for Proto-Aymaran, the proto-language of a neighboring language family whose members are otherwise structurally very similar to Proto-Quechuan, and which has therefore long been considered an obvious candidate for deep shared ancestry. Some of these elements are explainable as phonesthemes, but most appear to reflect archaic Quechuan morphology. These findings are consistent with an emerging picture of the early Quechuan-Aymaran contact relationship in which Quechuan structure was reformatted on the Aymaran template.